Spaces:
Sleeping
Sleeping
File size: 3,822 Bytes
944aab6 3f62146 944aab6 3f62146 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
retriverText = """ This microservice integrates with the vector database to retrieve semantically relevant documents,\
with optional reranking for precision, ready for seamless use in ChaBo RAG workflows.
# Retriever and Reranker Microservice on Hugging Face Spaces
[ChaBo_Retrieval](https://huggingface.co/spaces/GIZ/chatfed_retriever0.3) hosts a Retrieval and Reranker mciroservice.\
Some of key feature of Retrieval service are:
- The embedding of the user query is done by retriever itself using Sentence-Transformer.
- ReRanker is available as optional component.
- This is rate determining step as the emedding of user query can be compute intensive if using dedicated model.
- Model config, Qdrant server url and other params can be set through \
[params.cfg](https://huggingface.co/spaces/GIZ/chatfed_retriever0.3/blob/main/params.cfg)
```
[vectorstore]
# Qdrant-Server usage:
PROVIDER = qdrant
URL = giz-chatfed-qdrantserver.hf.space
COLLECTION_NAME = EUDR
[embeddings]
MODEL_NAME = BAAI/bge-m3
[retriever]
TOP_K = 10
SCORE_THRESHOLD = 0.6
[reranker]
MODEL_NAME = BAAI/bge-reranker-v2-m3
TOP_K = 10
ENABLED = true
# use this to scale out the total docs retrieved prior to reranking (i.e. retriever top_k * TOP_K_SCALE_FACTOR)
TOP_K_SCALE_FACTOR = 2
```
**API documentation**: 1 API Endpoint
### api_name: /retrieve
Params:
- query(str): Required
- collection_name(str): collection_name in the Qdrant server which need to be queried. Defualts to None.
- filter_metadata(dict): metadata filtering for Qdrant vector store which will be
applied to the collection mentioned above. Defuals to None
Returns: List of retrieved context along with metadata as string,
where each context is dict with two key 'answer' and 'answer_metadata'
**How to Connect**
```python
from gradio_client import Client
# Replace with your actual Space URL (e.g., https://your-username-retriever_space.hf.space)
retriever_url = "https://giz-chatfed-retriever0-3.hf.space/"
client = Client(retriever_url)
result = client.predict(
query="What is Circular Economy",
collection_name="Humboldt",
filter_metadata=None,
api_name="/retrieve"
)
```
For more info on Retriever and code base visit the following links:
- ChaBo_Retriever : [**ReadMe**](https://huggingface.co/spaces/GIZ/chatfed_retriever0.3/blob/main/README.md)
- ChaBo_Retriever: [**Codebase**](https://huggingface.co/spaces/GIZ/chatfed_retriever0.3/tree/main)""" |